Search

The account of Classification and Regression Trees and allied methodologies is primarily a lead-in to the Chapter 3 account of the random forests (RF) approach, as implemented in the B randomForest b package. Tree-based methods potentially suffer, along with all predictive modelling methods, from what has come to be termed "algorithmic bias" - a bias that is, however, in the data used and not inherent in the algorithm. The observations that are not included in a particular bootstrap sample are referred to as the out-of-bag observations and used to compare predictions made by the tree with observed outcome values. [Extracted from the article]